After the NFL schedule was released last week, SI's Peter King published an insider's look into how the scheduling sausage gets made. The key takeaway: Out of the more than 500,000 scheduling possibilities, roughly 175 are put forth for further consideration. This begs the question, how does the league decide which of those 175 possibilities becomes the actual schedule?
Throughout the article, King gives bits and clues, such as atypically arduous road trips for any given team. But NFL senior vice president of broadcasting Howard Katz sums it up by saying, "This schedule has the best television without putting really any unfair burden on any club.'"
Ah, television. Of course. Television rules the NFL culturally and financially, so it makes sense that putting the best games on national broadcasts would be in the NFL's interest. But Katz's statement is purposely vague. What, exactly, is the "best" television?
This may sound like an academic question, but given how much team quality shifts year-to-year, it's incredibly hard to predict which games will be worth watching eight months before the season kicks off. One of the NFL's great claims, that it features the most parity of any professional sports league, is precisely what makes primetime so tricky to schedule. Consider the Texans: Coming off a 12-4 season with a combination of young talent and experienced veterans, they were given three 2013 primetime games. When the Texans followed up that impressive season with a total collapse, something the schedulers could never have foreseen, this meant viewers were stuck watching a 2-14 team for three primetime games. The schedule-makers can account for a lot of factors, but reversal of fortune is not one of them.
In fact, the NFL does seem to rely heavily on previous year record when scheduling primetime matchups. I looked at two different types of win percentages for teams on primetime since 2011: Their winning percentage the prior year and their "current" winning percentage, or their record at the time the game was played. This made clear that ESPN and NBC receive better than average games.
NBC has had the benefit of the best games, aided by the possibility of flexing games in the final weeks of the season -- they will have the option to do so weeks 5-10 this year as well -- with a prior year win percentage of 63 percent, or the same winning percentage as a 10-win team. ESPN's slate was slightly worse, with a 56 percent winning percentage (nine-win team). Unsurprisingly, NFL Network games were the least impressive, with a shade over 50 percent winning percentage the prior year. This is largely due to the league requirement that every team play in primetime at least once per year, so the league dumps the also-rans on Thursday rather than on its network partners.
Despite the league's efforts to give networks the best possible games, it doesn't play out quite as well as they would hope. If games were assigned completely randomly, we would see the "current" win percentage approach 50 percent. Interestingly, we see something close to this already, even without random assignment of games. This is because, although not random at all, the schedulers' lack of precognition negates some of their careful planning; they may think they're giving networks a great matchup, but there's so much uncertainty in team performance that they don't really know.
Even though NBC's matchups feature teams with a winning percentage the prior year of 63 percent, their actual winning percentage for the current season was just under 60 percent, so teams were a little worse than the prior year (but, still better than average). ESPN's regression was slighter, but brought their actual matchup win percentage to 54 percent, or that of an 8.6-win team over the course of a season. NFL Network actually saw a below league-average quality game, with a matchup win percentage of 47 percent. Going by this measure, Thursday night games have indeed been consistently less appealing than Sunday and Monday night games.
While this gives us a clearer picture of how networks are allocated games, what about the teams themselves? That is, how does the NFL decide which teams get to play in primetime? The answer is not as obvious as you think.
Given what we've found about primetime games and win percentage, my first thought was that more successful teams are put on primetime as often as possible the following year. To a very large extent, this proves correct.
This relationship looks almost perfect, except for the group of teams with the most primetime appearances (Cowboys, Giants, Eagles, Bears, Broncos, Steelers and Saints). It seems they're immune to the win percentage requirement. Perhaps this can be explained by random poor seasons -- the Steelers, Giants and Saints come to mind -- but there must be something more.
Perhaps larger markets are put on primetime more often in order to capture a larger viewership. But, when measuring market size using a 75-mile radius -- the same radius in which the NFL blackout rules apply -- this doesn't seem to be the case.
There is a slight positive correlation here, but it doesn't explain the win percentage outliers.
The Cowboys, Giants, Eagles, Bears, Broncos, Steelers and Saints (our outlier group from the win percentage relationship) are generally regarded as some of the league's most popular teams. Perhaps it's not market size but general popularity that determines who will be on primetime. To see if that is the case, I compared primetime appearances with official team page Facebook Likes independent of geography. Since primetime games are nationally televised, it would make sense that the league would use these slots to showcase teams with widespread popularity, not just large markets.
This does, indeed, seem to be the missing piece of the puzzle. The teams with the most Likes are exactly the ones that we couldn't explain via win percentage.
There's a bit of chicken-and-egg relationship at work here. Are teams on primetime more because they're popular, or are they popular because they're on primetime more? To answer this question, I ignored primetime appearances for the time being and just looked at the relationship between Facebook Likes (popularity) and win percentage. If these two variables bear a strong relationship, then causation will be pretty difficult to determine.
There's a pretty moderate correlation between Facebook Likes and win percentage, but it's not very strong, and significantly weaker than the correlation we saw between prior year win percentage/Facebook Likes and primetime appearances. To be fair, I didn't take into consideration historical success, which is commonly regarded as a significant determinant for popularity. At that point, we would get into a whole causational mess about whether primetime appearances, historical success, or popularity is the determining factor for the other two. But that's another question in itself.
The upshot: Good teams get put on primetime more often, but as long as a team is popular, short-to-medium term failure doesn't lead to a reduction of primetime appearances. This is how the Cowboys can have five primetime games in 2014, which is why I call this popularity correlation the Schadenfreude Effect.
All this means we should be very impressed with the NFL schedulers. Whenever something makes statistical sense, it means there's an identifiable pattern. A pattern suggests systematization, which means the schedule-makers are predictable and unbiased (remember, they don't determine opponents, just order). Aside from a mess of logistical and practical considerations, they clearly consider two key factors when allocating primetime appearances: a team's win percentage and its overall popularity. So get used to watching the Bears, Steelers and NFC East over and over. They're not getting less popular any time soon.